Journal of

Mechanical
Science and
Technology
Journal of Mechanical Science and Technology 23 (2009) 1620~1627
www.springerlink.com/content/1738-494x
DOI 10.1007/s12206-008-1106-1

Predicting performance of radial flow type impeller of

centrifugal pump using CFD†
Suthep Kaewnai, Manuspong Chamaoot and Somchai Wongwises*
Fluid Mechanics, Thermal Engineering and Multiphase Flow Research Lab. (FUTURE), Department of Mechanical
Engineering, King Mongkut’s University of Technology Thonburi, Bangmod, Bangkok 10140, Thailand

(Manuscript Received August 31, 2006; Revised April 20, 2008; Accepted November 17, 2008)
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Abstract

The main objective of this work is to use the computational fluid dynamics (CFD) technique in analyzing and pre-
dicting the performance of a radial flow-type impeller of centrifugal pump. The impeller analyzed is at the following
design condition: flow rate of 528 m3/hr; speed of 1450 rpm; and head of 20 m or specific speed (Ns) of 3033 1/min in
US-Units. The first stage involves the mesh generation and refinement on domain of the designed impeller. The second
stage deals with the identification of initial and boundary conditions of the mesh-equipped module. In the final stage,
various results are calculated and analyzed for factors affecting impeller performance. The results indicate that the total
head rise of the impeller at the design point is approximately 19.8 m. The loss coefficient of the impeller is 0.015 when
0.6 < Q/Qdesign < 1.2. Maximum hydraulic efficiency of impeller is 0.98 at Q/Qdesign = 0.7. Based on the comparison of
the theoretical head coefficient and static pressure rise coefficient between simulation results and experimental data,
from previous work reported in the literature [Guelich, Kreiselpumpen, Springer, Berlin, 2004], it is possible to use this
method to simulate the performance of a radial-flow type impeller of a centrifugal pump.
Keywords: Centrifugal pump; CFD; Impeller; Radial flow
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

number of researchers [1-8]. However there are few

1. Introduction
studies applying the CFD for predictions involving a
The use of CFD plays an important role in fluid radial flow type impeller. To the best of our knowl-
mechanics. Due to the progress of numerical methods edge to date, there have been only two works carried
and computer capability, the impeller design for cen- out by [9] and [10], in dealing with this issue.
trifugal pumps nowadays has been analyzed by using [9] analyzed about 30 pump impellers having spe-
3D-Navier stokes program or CFD software to predict cific speeds between 12 to 160 (metric units) using a
impeller performance in advance. However, both the commercial 3D-Navier Stokes program for a standard
prediction and design cannot be done easily without k-ε turbulence model. The head and efficiency ob-
sufficient information and experience. If the result tained from the software were compared to those
from the CFD is much different from the designed obtained from the measurement. A set of rather strin-
value, the impeller could be redesigned before being gent rules for the CFD calculations was developed in
put into real production. order to reduce the effect of grid and calculation pa-
Over the years, the application of CFD in fluid me- rameters on the results.
chanics and turbomachinery has been studied by a [10] performed a 3D-CFD simulation of the impel-
†This paper was recommended for publication in revised form by Associate ler and volute of a centrifugal pump having a specific
*
Editor Seungbae Lee speed of 32 (metric units) and an outside impeller
Corresponding author. Tel.: +662 470 9115, Fax.: +662 470 9111
E-mail address: somchai.won@kmutt.ac.th
radius of 204.2 mm, by using CFX codes. A 3D flow
© KSME & Springer 2009 simulation for the impeller with a structured grid was
 S. Kaewnai et al. / Journal of Mechanical Science and Technology 23 (2009) 1620~1627 1621

presented. A sensitivity analysis was done based on

3. Mathematical model basic equation
grid quality and various turbulence model. The final
suitable impeller model was used for a 3D quasi- In the present study the commercial CFD software
unsteady flow simulation of the impeller-volute stage. CFX 5.5 was used. For three-dimensional incom-
The flow simulation was performed for several impel- pressible and unsteady flow, the continuity equation
ler blades and volute tongue relative positions. and momentum equation in the rotating coordinate
Although some information is currently available system are as follows:
on applying the CFD for predicting of radial flow Continuity equation
type impeller, there still remains room for further
research. In the present study, the main concern is to ∂ρ ur ur
+ ∇. ρV = 0 (1)
- predict theoretical head or theoretical head coeffi- ∂t
cient at any volumetric flow rate
- predict total head rise or total head coefficient of Momentum equation
impeller at any volumetric flow rate
ur
- predict impeller loss or impeller loss coefficient dV ur ur ur ur ur ur ur r
at any volumetric flow rate ρ
dt
( ) (
+ ∇P = ρg + µ ∇ 2 .V − 2 ρΩ × V − ρ Ω × Ω × r )
- determine effect of turbulence model on total (2)
head rise of impeller
- examine uniformity of Cm2 distribution which af-
4. Parameters in calculation of performance
fects the mixing losses (Q-H curve instability)
- elucidate the method to validate the CFD results In analyzing the impeller with CFD code, velocity
and pressure at the control surface on the upstream
and downstream side of the impeller can be integrated
2. Impeller design process as follows:
Usually, there are two methods to design the impel- • Static pressure rise coefficient of the impeller
ler: direct method and inverse method. In the present
study, the direct method is used. Fig. 1 shows the
impeller design process which starts with specifica-
ψP =
2
ρU 22
(
P2 static − P1static ) (3)

tions of the impeller by identifying flow rate, Q head,

H and speed, n. • Total head rise coefficient

ψ LA =
2
ρU 22
(
P2total − P1total ) (4)

Fig. 1. Design process of impeller. Fig. 2. 3D visualization of impeller.

1622 S. Kaewnai et al. / Journal of Mechanical Science and Technology 23 (2009) 1620~1627

Table 1. Simulation.

PARAMETERS CFX 5.5

Flow simulation domain Single impeller flow channel

Mesh Structured

Fluid Water at 25 °C

Inlet Total pressure = 101325 Pa

Mass flow rate = variable
Outlet
(kg/s)
Fig. 3. Mesh on simulation domain. Wall No slip

Turbulence model k - ε , k - ω, RNG k - ε

• Theoretical head coefficient Turbulence intensity 1% , 5% , 10%
Maximum residual
10-4 (RMS)
⎛ UC UC u1 ⎞⎟ 2 gH th convergence
ψ th = 2⎜ 2u 2 − = (5)
⎜ U U 22 ⎟⎠ U 22
⎝ 2
Table 2. Mesh details.

• Hydraulic loss or loss coefficient of the impeller Number of Number of Minimum

Maximum
Mesh edge length
node element face angle
ratio
2 gZ La
ζ LA = ψth − ψ LA = (6) A 17340 14720 18.2653 68.7239
U 22
B 25080 21300 18.0644 71.7829
• Hydraulic efficiency of the impeller C 34200 29820 18.0644 51.6179

D 43320 38340 18.0644 58.7055

ψ
ηh,impeller = LA (7) E 50440 44160 17.5859 88.2421
ψ th
Total head rise F 58200 51520 17.5859 71.8257
or ηh,impeller =
Theoretical head G 65960 58880 17.5859 64.5811

H 81480 73600 17.5859 50.8711

5. Analysis of flow in the impeller I 91500 81900 17.4755 90.5938

The impeller analysis for the radial flow centrifugal

pump was based on these design details: speed of
1450 rpm; flow rate of 528 m3/hr; head of 20 m or
specific speed of 3033 1/min in US units. The dimen-
sion of the impeller is obtained as follows: impeller
diameter, 300 mm; exit width, 42 mm; total number
of blade impellers, 7; and angle at the exit, 24.5°. The
impeller is as shown in Fig. 2. The analysis starts with
the mesh generation and refinement on the domain.
This mesh has a hexahedral shape as shown in Fig. 3.
The analyzed parameters are shown in Table 1.
6. Results and discussion
6.1 Effect of number of mesh
Since the number of mesh nodes or openings af-
fects the analysis, a suitable mesh number should be
determined. Generally, a value of about 45,000 –
75,000 nodes is used in impeller analysis. In this work,
nine models of mesh were constructed on domain to
determine the effect from the mesh number, as shown
in Table 2.
The analysis of flow in the impeller indicates that a
mesh number in the range of 45,000 – 75,000 nodes
yields similar results. However, to facilitate further
study, the E model mesh was selected for the next
analysis.
6.2 Effect of the change of the turbulence model
In CFD analysis, turbulence model should be se-
lected. In this work, turbulence models (k - ε , k - ω ,
 S. Kaewnai et al. / Journal of Mechanical Science and Technology 23 (2009) 1620~1627 1623

Table 3. Performance parameters obtained from CFX 5.5.

Parameter ψp ψth ψLA ζ LA

1% 0.1811 0.2107 0.1980 0.0127
Turbulence
5% 0.1810 0.2107 0.1979 0.0128
intensity
10% 0.1805 0.2106 0.1974 0.0132
0 0.1809 0.2107 0.1979 0.0128

Surface 1E-4 0.1810 0.2108 0.1980 0.0128

roughness 1E-3 0.1812 0.2115 0.1984 0.0131
(mm) 0.01 0.1824 0.2163 0.2013 0.0150
0.1 0.1834 0.2284 0.2068 0.0215

Fig. 4. Performance of impeller obtained from nine models of

mesh.

Fig. 6. Total head rise of impeller obtained from various turbu-

lence intensity.

value determined at the design point is 20 m. Hence,

Fig. 5. Total head rise of impeller obtained from various turbu-
lence models.
RNG k - ε) are selected to compare the use of various
turbulence models on impeller performance. The
analysis reveals that the head values gained from all
three turbulence models are similar, with only 0.3%
difference. However, the turbulence model k - ε is
selected for further analysis.
6.3 Effect from the change of the turbulence inten-
sity
The change in turbulence intensity affects the im-
peller performance. In this work, three levels of turbu-
lence intensity are used: 1%, 5%, and 10%. The
analysis reveals that the head values gained from all
three turbulence intensity levels are similar; however,
5 % was selected for further analysis.
6.4 Impeller performance
From this study, the performance of a impeller
could be determined as shown in Figs. 7-10. The head
it could be seen from Fig. 7 that the total head rise
value of 19.8 m is nearly to the designed head. Fig. 8
shows the static pressure rise coefficient and theoreti-
cal head coefficient at different volumetric flow rates
of the impeller. In real practice, it is quite difficult to
measure the static pressure rise in experiments. From
Fig. 9, it can be seen that a small amount of loss oc-
curred in the impeller at the Q/Qdesign value within the
range of 0.6-1.2. The loss coefficient at this range is
around 0.015. This shows the high impeller efficiency
and no recirculation. The performance of impeller
decreases outside this range. At low volumetric flow
rate, there is recirculation of water at the impeller
inlet, starting at the Q/Qdesign value below 0.6. The
hydraulic efficiency of the impeller is the ratio of total
head coefficient and theoretical head coefficient,
ψ
ηh = LA . From Fig. 10, the best hydraulic effi-
ψth
ciency value is at Q/Qdesign = 0.7. The use of CFD
software makes the hydraulic efficiency of the impel-
ler prediction easier. Figs. 11-12 show meridian ve-
locity at the impeller exit from hub to shroud. The
1624 S. Kaewnai et al. / Journal of Mechanical Science and Technology 23 (2009) 1620~1627

Fig. 7. Plot of total head rise against volumetric flow rate. Fig. 10. Plot of hydraulic efficiency of the impeller against ratio
of volumetric flow rate.

Fig. 8. Plot of static pressure coefficient and theoretical head

coefficient against ratio of volumetric flow rate.

Fig. 11. Plot of meridian velocity at impeller exit.

Fig. 9. Plot of loss coefficient against ratio of volumetric flow

rate.

non-uniformity of the meridian velocity distribution

affects the mixing losses and Q-H curve instability at
low volumetric flow rate. This is unavoidable since
the impeller revolved and produced friction at its
surface. It can be seen from Table 3 that losses in the
impeller increase with increasing surface roughness. Fig. 12. Plot of meridian velocity at impeller exit in 3D.
 S. Kaewnai et al. / Journal of Mechanical Science and Technology 23 (2009) 1620~1627 1625

The theoretical head coefficient obtained can be com-

7. Validation of CFD results
pared with that obtained from Eq. (5).
Simulation results can be validated by comparison In the present study, the experimental work was not
with the experimental data. In this work, static pres- carried on. But, results from a previous study reported
sure rise coefficient and theoretical head coefficient in the literature [11] show good agreement between
are the parameters that can be used for verification. the theoretical head coefficient and the static pressure
The static pressure rise can be obtained from the rise coefficient. However, due to recirculation,
static pressure data taken from the pressure transducer agreement becomes worst at very low (Q/Qdesign). An
installed on the surface of impeller at the inlet and example of the comparison obtained from [11] is
outlet of pump. Then, the static pressure rise coeffi- shown in Fig. 13.
2 gH p
cient ψ p,exp calculated from ψ p ,exp = is
U 22 8. Conclusion
compared with that obtained from Eq. (3). In the impeller analysis with CFD software, it was
The other parameter used to verify the simulation found that the suitable mesh number was more than
results is the theoretical head coefficient. The first 50,000 nodes. The turbulence model and turbulence
step to derive this parameter starts from the hydraulic intensity yielded similar results at the same condition.
efficiency of pump, ηh,pump, obtained from The prediction of impeller performance indicated that
the loss coefficient was minimum and the hydraulic
QspQE Qh efficiency was maximum at a volumetric flow rate
1+ ++
Q Q Q Q/Qdesign = 0.7. The total head rise values gained from
ηh , pump = (8)
1
+
∑ PRR + ∑
Ps 3 + Pm + Per the analysis were acceptably non-significantly differ-
ent from the head at the design point. The surface
η Pw roughness value was found to have a high effect on
loss. That is, at higher surface roughness, the value of
where the water horse power, Pw is calculated from the loss coefficient was also found to increase. The
loss coefficient of the impeller is 0.015 when 0.6 <
Pw = ρgQH tot (9) Q/Qdesign < 1.2. Maximum hydraulic efficiency of
impeller is 0.98 at Q/Qdesign = 0.7.
where η, H tot and Q are obtained from the perform-
ance curve of the pump. Acknowledgments
Then, the theoretical head, H tot , can be determined
H tot The present study was supported financially by the
from η h , pump = and the theoretical head coef- Thailand Research Fund (TRF). Their guidance and
H th
assistance are gratefully acknowledged. The authors
2gH tot would like to express their appreciation to Professor J.
ficient from ψ th,exp =
ηh, pumpU 22 F. Guelich, of the Laboratory for Hydraulic Machines,
Ecole Polytechnique Federale de Lausanne, Switzer-
land, for his valuable comments and suggestions.

Fig. 13. Comparison between results obtained from experiment
and CFD [11].
Nomenclature-----------------------------------------------------------
B : Width at the exit of the impeller (mm.)
Cm2 : Meridian component of absolute velocity at exit
(m/s)
CU1 : Circumferential component of absolute velocity
at inlet (m/s)
CU2 : Circumferential component of absolute velocity
at exit (m/s)
FEA : Finite element analysis
g : Gravitational acceleration (m/s2)
H : Design head of the impeller (m)
1626 S. Kaewnai et al. / Journal of Mechanical Science and Technology 23 (2009) 1620~1627

Hp : Measured static pressure head (m) References

Hth : Theoretical head (m)
Htot : Total head of pump (m) [1] P. Cooper and E. Graf, Computational fluid dy-
k : Turbulence kinetic energy namical analysis of complex internal flows in cen-
l : Circumference of the impeller exit on rotational trifugal pump, Proceedings of the 11th International
polar coordinate for each blade impeller (mm.) Pump Users Symposium, Houston, USA, (1994),
Ns : Specific speed (1/min) 83-93.
P : Pressure (N/m2) [2] FC. Visser, Some user experience demonstrating the
Per : Power losses by component of axial thrust bal- use of CFD for cavitation analysis and head predic-
ance device (kW) tion of centrifugal pump, Proceedings of the ASME
Pm : Mechanical power losses (kW) Fluids Engineering Division Summer Meeting
PRR : Disk friction power (kW) (FEDSM’01), New Orleans, La, USA, May-June
Ps3 : Power losses dissipated in inter-stage seal (kW) (2001), paper FEDSM2001-18087.
Pw : Water horse power (kW) [3] E. Blanco-Marigorta, J. Fernandez-Francos, J. L.
P1static : Static pressure at the impeller inlet (N/m2) Parrondo-Gayo and C. Santolaria-Marros, Numeri-
P2static : Static pressure at the impeller exit (N/m2) cal simulation of centrifugal pumps, Proceedings of
P1total : Total pressure at the impeller inlet (N/m2) the ASME Fluids Engineering Division Summer
P2total : Total pressure at the impeller exit (N/m2) Meeting (FEDSM’00), Boston, Mass, USA, June
Q : Any volumetric flow rate (m3/s) (2000), paper FEDSM00- 11162.
Qdesign : Volumetric flow rate at the design point (m3/s) [4] J. Gonzalez, J. Fernandez-Francos, E. Blanco and C.
QE : Leakage flow through axial thrust balance de- Santolaria-Marros, Numerical simulation of the dy-
vice (m3/s) namic effects due to impeller-volute interaction in a
Qh : Leakage flow through auxiliaries (m3/s) centrifugal pump,Transactions of the ASME, Jour-
Qsp : Leakage flow through seal at impeller inlet (m3/s) nal of Fluids Engineering, 124 (2002) 348-355.
RNG : Renormalization group [5] J. L. Parrondo-Gayo, J. Gonzalez-Perez and J. Fer-
r : Radius from the reference point (m) nandez-Francos, The effect of the operating point on
U : Circumferential velocity (m/s) the pressure fluctuations at the blade passage fre-
U2 : Impeller tip speed (m/s) quency in the volute of a centrifugal pump, Trans-
V : Velocity vector (m/s) actions of the ASME, Journal of Fluids Engineering,
y : Any distance along the circumference of the 124 (2002) 784-794.
exit measured on rotational polar coordinate [6] F. Gu, A. Engeda, M. Cave and J. L. Di Liberti, A
(mm.) numerical investigation on the volute/diffuser inter-
z : Any distance along the width at the exit of im- action due to the axial distortion at the impeller exit,
peller measured from Hub to Shroud (mm.) Transactions of the ASME, Journal of Fluids Engi-
ZLa : Hydraulic losses of the impeller neering, 123 (2001) 475-483.
η : Efficiency of pump [7] F. Muggli, P. Holbein and P. Dupont, CFD calcula-
ηh : Hydraulic efficiency of the impeller tion of a mixed flow pump characteristic from shut-
µ : Absolute viscosity (kg/m-s) off to maximum flow, Proceedings of the ASME
ρ : Density of medium (kg/m3) Fluids Engineering Division Summer Meeting
Ω : Angular velocity (rad/s) (FEDSM’01), New Orleans, La, USA, May-June
ε : Turbulence eddy dissipation (2001), paper FEDSM2001-18072.
ω : Turbulent frequency [8] C. Cravero and M. Marini, Modeling of incom-
ψ : Head coefficient pressible three-dimensional flow in rotating turbo-
ψp : Static pressure rise coefficient of the impeller machinery passages, Proceedings of the ASME
ψLA : Total head rise coefficient of the impeller Fluids Engineering Division Summer Meeting
ψth : Theoretical head coefficient (FEDSM’02), Montreal, Quebec, Canada, July
ζLA : Impeller loss coefficient (2002), paper FEDSM2002-31177.
→ ∂ ∂ ∂ [9] J. F. Guelich, J. N. Favre and K. Denus, An assess-
∇ : Operator ≡ i + j +k
∂x ∂y ∂z ment of pump impeller performance predictions by
3D-Navier Stokes calculations, Proceedings of the
 S. Kaewnai et al. / Journal of Mechanical Science and Technology 23 (2009) 1620~1627 1627

ASME Fluids Engineering Division Summer Meet- Suthep Kaewnai obtained a B.

ing (FEDSM’97), Vancouver, Canada, June 1997, S. degree in Mechanical Engi-
paper FEDSM1997-3341. neering, 1980 from the King
[10] M. Asueje, F. Bakir, S. Kouidri, F. Kenyery and R. Monkut’s University of Tech-
Rey, Numerical Modelisation of the Flow in Cen- nology Thonburi and M. S.
trifugal Pump: Volute Influence in Velocity and degree in Mechanical Engineer-
Pressure Fields, International Journal of Rotating ing, 1983 from Chulalongkorn
Machinery, 3 (2005) 244-255. University. He is currently an
[11] J. F. Guelich, Kreiselpumpen, Springer, Berlin, assistant professor at King Mongkut’s University of
2004 (in German). Technology Thonburi. Suthep’s research interests are
in the area of pumps and small hydroturbine.

Somchai Wongwises is cur- Manuspong Chamaoot received

rently a Professor of Mechani-
cal Engineering at King Mong-
kut’s University of Technology
Thonburi, Bangmod, Thailand.
He received his Doktor In-
genieur (Dr.Ing.) in Mechanical
Engineering from the Univer-
sity of Hannover, Germany, in 1994. His research
interests include two-phase flow, heat transfer en-
hancement, and thermal system design. Professor
Wongwises is the head of the Fluid Mechanics,
Thermal Engineering and Two-Phase Flow Research
Laboratory (FUTURE).
a B. S. degree ,1972 and M.S.
degree in Mechanical Engineer-
ing, 1979 from the King
Monkut’s University of Tech-
nology Thonburi. He is cur-
rently an assistant professor at
King Monkut’s University of
Technology Thonburi. His research interests are in
the field of mechanical vibration for rotating equip-
ment and computational fluid dynamics.